try {} finally {}

[Tony Kimball]

I have seen enough instances of this behaviour so that I am fairly
certain that the bug I am about to describe exists; however, I have
not been able to produce a small case.  This is probably the same
bug that Martin Kahlert was trying to isolate.  I hope that I can
contribute a somewhat better characterization of the bug, if nothing else:

When there is a try {} finally {} block, with no catch clause,
sometimes exception.cc:get_ttype_entry will return null.  This results
in a segv in _Jv_IsAssignableFrom during a throw, which sometimes will
recursively segv in the handler.  

I am using the 3.0.1 branch.